In this paper reconfigurable robust Linear Quadratic Regulator (LQR) and Kalman filter (KF) are developed for discrete-time systems subjected to faults, explicitly taking into account uncertainty both in the model of the system and in the estimated faults from the fault detection and diagnosis (FDD) part. For each separate actuator (sensor) with which the system is robustly stabilizable (detectable), a robust LQR gain (robust KF gain) is designed. After each occurrence of faults, a reconfiguration of the LQR (KF) is performed by an ppropriate mixing of the pre-designed gains, resulting in the optimal obust LQR (KF) for the current faulty system. The approach is computationally attractive and can handle sensor, actuator and component faults. The approach is tested on an industrial actuator benchmark model.